Keep Your Codebase Yours: Working with AI Agents on Long-Lived Software

A free 60-minute pilot workshop for senior engineers

Keep Your Codebase Yours: Working with AI Agents on Long-Lived Software

You’ve invested years in building a codebase with real structure: clear boundaries, deliberate APIs, and carefully considered architecture. Now you’re leveraging AI agents for development, code has started arriving faster than anyone can understand it, and the design you protected feels at risk.

This workshop shows you how to get the speed of AI-assisted development without giving up the coherence that keeps a codebase dependable and well-understood. We’ll use a prepared example to walk through the judgment calls together:

  • which work to hand to an agent, and which design decisions stay with you
  • how to plan before the agent writes code, so the result fits your architecture
  • how to review agent output for design drift, not just bugs
  • how to keep ownership, with code you understand, can explain, and would ship

We’ll explore this in the context of a Mac app, but the principles apply to any large, long-lived codebase. You’ll leave with techniques you can bring back to your own code and team.

Troy Stephens

Your instructor

Troy Stephens

I spent 16 years on Apple’s AppKit team, helping build macOS developer functionality that underpins apps from Safari to Keynote, and designing and presenting 12 WWDC talks along the way. I’ve built solid indie Mac apps that have survived years of OS changes, and I’ve helped integrate AI dev workflows into shipping of a major app — building experience with Claude Code, ChatGPT Codex, Grok Build, GrokBot, and CodeRabbit. I’ll show you how to maintain control over your codebase while bringing your own apps into the future.

  • 16 years on Apple’s AppKit team
  • 12 WWDC talks
  • Shipped a major app using LLM-assisted development